Should Stronger Keywords Subsidize Weaker Keywords?

Some keywords perform great. Some keywords perform OK, and some keywords stink. Today I’d like to write about keyword portfolio economics, asking, “Does it ever make sense to use stronger keywords to subsidize weaker keywords?” My answer is “Yes—when done correctly.”

Efficiently bidding a large keyword portfolio is not simple. A well-built keyword portfolio includes many terms with low traffic. Bidding such long tail terms effectively requires sophisticated math. We won’t dive into the equations today (cheers from the bleachers), but instead consider the simplified problem of deciding when to turn off a marginal keyword segment.

Let’s suppose you’re managing search for an online retailer and your goal is to maximize sales while spending not more than 35 cents of revenue on advertising. That is, you’re seeking a 35% A/S target (advertising to sales ratio). Further, let’s suppose your cost-of-goods-sold is 50% and you have variable sales expenses of 10% (pick-pack-ship, credit card discount fees, phone support, etc).

Here’s the simplified problem. Let’s suppose you’ve tested eight campaigns or keyword performance segments, imaginatively named “A” through “H”, and you’ve reached a pretty good understanding of how each segment performs. This table gives monthly clicks, CPCs, and SPCs (sales-per-click), for each segment.

For this toy example, let’s pretend the keywords in each segment have reasonably similar economics.

To focus on the subsidization argument, let’s make a huge and critical assumption that all the keywords are being bid intelligently. That is, for less efficient terms, you’ve already experimented with bidding them down the page, and found that the traffic drops off to nothing at your efficiency needs, so you truly only have the option to turn them on inefficiently on the first page or not have them generate much traffic. And for the super-efficient terms in bucket “A,” you’ve already bid them to top position, so there’s no place higher for them to go.

Here are those campaign options again, presented graphically:

Paid search inventory often looks like that—a modest amount of good solid traffic surrounded by oceans of lower-quality traffic.

OK. We said your goal was 35% A/S, so you’d run bundles “A,” “B,” and “C,” yes? In the bar chart, take the green bars and skip the red, right?

This table presents the same data as the first, but also provides cumulative results.

It shows that running bundles A, B, and C, at A/S ratios of 3%, 22% and 35%, respectively, produces a combined campaign running at an overall 19% A/S. In dollar terms, those three segments combined provide 35K clicks, $15k in cost, $80k in sales – and that’s a 19% A/S.

Efficient advertising is good, but we said our goal in this example was maximizing sales while not spending more than 35% of revenue on advertising. With those goals, 19% is too efficient – we’re leaving top line on the table.

We need additional sales, and segment “D” is where we’ll find them. If we picked up the entire segment, the additional 100K clicks at 60% A/S (ugh!) would torpedo our overall A/S, raising it up to 42% and destroying our efficiency budget. We can use some of “D” though. Taking the best words from “D” and adjusting bids appropriately, let’s assume we can pull another 52K clicks at $31k in cost yielding an additional $52K in sales. (We added that row into the cumulative results, italicized in light gray.)

Here’s a sales-cost tradeoff curve:

The black dot shows sales and costs for running bundles “A,” “B,” and “C”: $80K sales, $15K cost. The yellow dot shows sales and cost for running “A,” “B,” “C,” and the best of “D”: $132K sales and $46K in cost.

By running some lesser-performing keyword segments, we added $52K in top line while staying within our efficiency budget. We’ve used great performance from strong keywords to subsidize poorer performance from weaker keywords.

Earlier we said “we need additional sales, and D is where we’ll find them.” That assumes a comprehensive and fully-built out keyword list. Before considering the best of the marginal terms, be sure you’ve not overlooked any undiscovered pockets of high performing terms. Very large high-quality keyword lists require lots of effort to create, manage copy for, and bid well—but the additional effort and complexity are worth it. Also be sure too that your technology never underbids any of the high-performing terms in higher segments.

So, is intra-portfolio subsidization a good thing? That depends. In this toy example, subsidization grew revenue within our efficiency target, which was our objective. By that measure, yes, it is a good thing.

But note this subsidization reduced profit. To put it another way: this subsidization invested bottom line to buy top line. The “A-B-C” scenario generates $17K in marketing contribution (assuming 50% COGs and 10% variable). The “A-B-C-best-of-D” scenario generates $6,600 in profit, a 62% reduction.

Why did profit go down? By setting a 35% A/S target, this retailer choose to favor top line sales over profit maximization. One formula for maximizing PPC profit suggests you’ll maximize profit contribution (in dollars, not percentages – you don’t put percentages in the bank) by spending half of your effective profit margin in percentage terms on marketing, keeping the other half as contribution towards profit. With 50% COGs and 10% variable, that implies the most profitable A/S target in this case is (1 – 0.5 – 0.1)/2, or 20%.

The A/S target is an advertiser’s choice, reflecting a need for profit vs. a need for sales. Choosing 35% isn’t “wrong” and choosing 20% isn’t “right”. Choosing an A/S target – equivalently, deciding the aggressiveness of a direct retailer’s marketing spend – is a high-level strategic decision involving considerations of growth, cash, and profit.

So, sometimes it might make sense to run some carefully selected poorer-performing keywords to buy additional top line. But it never makes sense to run horrible terms at any significant volume. In this scenario, it would never make sense to run “E,” “F,” “G,” or (horrors!) “H” terms. Those segments are just too dismal for direct marketers. (Brand marketers, they’re all yours if you want ‘em.) As mentioned earlier, to simplify this example we’re assuming that lowering bids on these marginal terms isn’t much of an option, assuming that dropping bids to the level needed to make the economics would require going so far down in the rankings that traffic would effectively vanish, which is essentially the same as turning them off.

We recommend auditing your PPC campaigns at the keyword level regularly. Should you observe high-volume high-cost keywords consistently performing worse than your efficiency target, dig in and start asking hard questions. If they’re not selling, why are high volume terms being bid aggressively? Is there economic justification? Or are those bids simply wrong?

Done improperly, subsidization can hide many PPC sins, obscuring bad bidding and allowing brand term performance to hide poor non-brand campaigns. (See point #3, “success means growing sales and profits from non-brand phrases”.) Done properly, lesser performing keywords offer some potential to grow sales, but of course this benefit comes at a cost. A good bid management system should hit any non-brand efficiency target you want, but there is a tradeoff between top and bottom line. The goal is not only efficiency, but volume too.

So back to where we started: Does it ever make sense to use stronger keywords to subsidize weaker keywords? Yes, when done carefully.

Alan Rimm-Kaufman leads the Rimm-Kaufman Group, a direct marketing services and consulting firm founded in 2003. The Paid Search column appears Tuesdays at Search Engine Land.